The Study Of Accuracy Variation Of MOD10A1 Snow Cover Datas In Rugged Terrain

The snow with high reflectivity and low thermal conductivity and snowmelt hydrological effect characteristics, is one of the most active natural elements in the earth’s surface. Most snow in China are mainly distributed in the high altitude mountainous area, such as Tibet Plateau, northern xinjiang and northeast region. On river basin scale, the mountain snow accumulation and ablation has an important influence on regional weather and water balance, ecological environment, thereby affecting the economic and social development of the region. Therefore, the accurate identification of the snow cover area is the primary basis about snow research, and has important research significance.The emergence and development of remote sensing technology promote the improvement of the efficiency and accuracy of snow cover mapping. MODIS snow cover products is currently the most widely used snow cover data. This article gets the "true value" of snow cover area from Landsat8 high-resolution remote sensing image by SNOWMAP method after the terrain correction, and selects total 27 period of MOD1OA1 snow area data as the research data in Heihe river upstream mountainous region of the northeast Tibet Plateau. Using the method of spatial statistical analysis respectively it evaluates the overall accuracy on the snow cover daily data (SCA) and the fractional of snow cover data (FSC) of MOD10A1. At the same time, the influence of different factors on accuracy of fractional snow cover data of MOD10A1 in rugged terrain, the spatial distribution of high-accuracy data and the relationship between factors and accuracy are our main research contents.Our study gets several conclusions:the snow cover area can be a good representation of the real value that gets from Landsat8 high-resolution remote sensing image using SNOWMAP method after the topographic correction; under clear-sky conditions in our study region, for snow cover daily data, the mean overall accuracy is less than 80%, the mean kappa is 0.4, MOD10A1 snow cover map and Landsat8 snow cover map have a general agreement; after analyzing 27 periods fractional snow cover data by synthesis,the mean absolute error is about 20%, the correlation coefficient is 0.67 with the data of "true value", so the accuracy of MOD10A1 snow cover data is higher in this region. The study also indicate that:MOD10A1 FSC value has a greater impact on the accuracy of the MOD10A1 fractional snow cover data, the greater the FSC value, the smaller the error, when the FSC value higher than 50%, the expectation and variance of relative error decrease obviously; when the vegetation coverage is low, the pixels of high accuracy have a high concentration degree, when the type of land is one of water, glaciers, tundra, meadow and bare land,the MOD10A1 fractional snow cover data has a high accuracy, but when the type of land is one of coniferous forest, shrubs, grassland, farmland and town, the MOD10A1 fractional snow cover data has a relatively low accuracy; high altitude has a more obviously impact on accuracy of MOD10A1 fractional snow cover data than low altitude, with the altitude increasing, accuracy gradually gets better, but up an altitude of 4000 m, the accuracy is stable; it shows a simple negative change between slope and accuracy of MOD10A1 fractional snow cover data, especially when the slope is greater than 35°,the accuracy is falling seriously, but in flat region, the accuracy is higher. Meanwhile, the relationship of between factors and the accuracy of MOD10A1 fractional snow cover data is further confirmed by analyzing spatial distribution of high-accuracy data, it provides a judgment for choosing effective data and forecast accuracy.